Automated generation of improved list-type answers in question answering systems

The invention relates to the field of Question Answering (QA) systems, particularly those based on machine learning models.

QA is a discipline within the field of computer science that aims to develop software systems capable of understanding and responding to natural language questions posed by users. The objective of QA systems is to retrieve the most relevant answer(s) to a given question from a large corpus of knowledge.

Over the past few decades, there have been significant advancements in natural language processing (NLP) and machine learning in general, which have greatly improved the performance of QA systems. These systems are now able to process large amounts of text data and extract relevant information to provide accurate and timely answers to users' questions.

QA systems based on machine learning models (typically, language models) are a popular approach. These models use various machine learning techniques to learn patterns and relationships in text data, allowing them to understand and interpret natural language more accurately. Machine learning-based QA systems are also able to improve their performance over time as they are exposed to more data and can adapt to changes in the underlying language model.

Some QA systems are capable of providing an answer that includes a list of items, such as a list of restaurants in a particular location, a list of top-rated books in a specific genre, or a list of steps to fix a particular computer error. These types of QA systems are able to understand that the information in the answer should be formatted as a list, and display the list as the answer to the question in a relevant way.

Many of today's prominent Web search engines incorporate such QA capabilities. In addition to providing traditional search results (hyperlinks to Web pages that are deemed relevant to a user's query), these search engines sometimes also provide a direct answer to the user's query when the query is phrased as a question or is otherwise deemed suitable for such direct answering. Depending on the question, the search engine may provide a direct answer in the form of a paragraph, a list, a table, or the like. For example, when the query is “What are the five tallest mountain peaks in the world?,” a search engine may provide a direct answer in the form of a list of mountains ranked by their elevation, followed by traditional search results of hyperlinks to Web pages relevant to query.

The foregoing examples of the related art and limitations related therewith are intended to be illustrative and not exclusive. Other limitations of the related art will become apparent to those of skill in the art upon a reading of the specification and a study of the figures.

The following embodiments and aspects thereof are described and illustrated in conjunction with systems, tools and methods which are meant to be exemplary and illustrative, not limiting in scope.

One embodiment relates to a computer-implemented method comprising, automatically: receiving data comprising: a question provided by a user, a list that constitutes a direct answer to the question, and an introductory text to the list; using a first machine learning model to classify the introductory text as redundant or nonredundant, based on the data; using a second machine learning model to classify the list as belonging to a certain list type out of multiple list types, based on the list; and providing to the user: (a) the introductory text, only if the introductory text has been classified as nonredundant, (b) all or only a subset of the items of the list, (c) an indication as to the number of non-provided items of the list or the number of all items of the list, if only a subset of the items is being provided in (b), and (d) a description of the certain list type.

Another embodiment relates to a system comprising: (i) at least one hardware processor; and (ii) a non-transitory computer-readable storage medium having program code embodied therewith, the program code executable by said at least one hardware processor to, automatically: receive data comprising: a question provided by a user, a list that constitutes a direct answer to the question, and an introductory text to the list; use a first machine learning model to classify the introductory text as redundant or nonredundant, based on the data; use a second machine learning model to classify the list as belonging to a certain list type out of multiple list types, based on the list; and provide to the user: (a) the introductory text, only if the introductory text has been classified as nonredundant, (b) all or only a subset of the items of the list, (c) an indication as to the number of non-provided items of the list or the number of all items of the list, if only a subset of the items is being provided in (b), and (d) a description of the certain list type.

A further embodiment relates to a computer program product comprising a non-transitory computer-readable storage medium having program code embodied therewith, the program code executable by at least one hardware processor to, automatically: receive data comprising: a question provided by a user, a list that constitutes a direct answer to the question, and an introductory text to the list; use a first machine learning model to classify the introductory text as redundant or nonredundant, based on the data; use a second machine learning model to classify the list as belonging to a certain list type out of multiple list types, based on the list; and provide to the user: (a) the introductory text, only if the introductory text has been classified as nonredundant, (b) all or only a subset of the items of the list, (c) an indication as to the number of non-provided items of the list or the number of all items of the list, if only a subset of the items is being provided in (b), and (d) a description of the certain list type.

In some embodiments, redundant introductory text is introductory text that conveys information which is included, explicitly or implicitly, in at least one of: the question and the list; and nonredundant introductory text is introductory text that conveys information which is not included, explicitly or implicitly, in at least one of: the question and the list.

In some embodiments, the method further comprises, or the program code is further executable to, train the first machine learning model by: obtaining multiple samples, each comprising a question, a list, and an introductory text; for each of the samples: computing a similarity measure between (i) at least one of: the respective question and the respective list, and (ii) the respective introductory text, and weakly labeling the respective introductory text as redundant or nonredundant, based on the similarity measure associated with the respective introductory text and on a predetermined similarity measure threshold; and adapting a language model based on the samples and the weak labels of the introductory texts.

In some embodiments, the method further comprises, or the program code is further executable to, train the second machine learning model by: defining each of multiple question patterns as corresponding to one of the multiple list types; obtaining multiple samples, each comprising a question, an introductory text, and a list; for each of the samples: matching the respective question with one of the question patterns, and weakly labeling the respective list and the respective introductory text as belonging to the list type which corresponds to the matched question pattern; and adapting a language model based on the samples and the weak labels of the lists and of the introductory texts.

In some embodiments, the method further comprises, or the program code is further executable to, train the second machine learning model by: defining each of multiple introductory text patterns as corresponding to one of the multiple list types; obtaining multiple samples, each comprising an introductory text and a list; for each of the samples: matching the respective introductory text with one of the introductory text patterns, and weakly labeling the respective list as belonging to the list type which corresponds to the matched introductory text pattern; and adapting a language model based on the samples and the weak labels of the lists.

In some embodiments, the multiple list types comprise at least some of: a sequence, which is a list of interdependent items with a meaningful order; a ranking, which is a list of independent items with a meaningful order; a catalog, which is a list of independent items without a meaningful order, wherein each of the items is an optional answer to the question; and an itemization, which is a list of independent items without a meaningful order, wherein all the items are necessary to answer the question.

In some embodiments, when only a subset of the items is being provided in (b), the method further comprises providing to the user a hyperlink to a document which contains all the items; and when all the items are being provided in (b), the method further comprises providing to the user an indication that all the items are being provided.

In addition to the exemplary aspects and embodiments described above, further aspects and embodiments will become apparent by reference to the figures and by study of the following detailed description.

Disclosed herein are a computer-implemented method, a system, and a computer program product that, given (a) a question provided by a user, (b) a list that constitutes a direct answer to the question, and (c) an introductory text to the list, employ machine learning models to automatically compile an improved answer to the question and provide it to the user.

The list and introductory text may be provided by a conventional machine learning-based QA system, as known in the art.

The Compilation of the Improved Answer May Include, for Example:

Reference is now made to, which shows a block diagram of an exemplary computing environment, containing an example of an environment for the execution of at least some of the computer code involved in performing the inventive methods, such as a modulefor automatically compiling an improved answer to a question and providing it to the user. In addition to module, computing environmentincludes, for example, a computer, a wide area network (WAN), an end user device (EUD), a remote server, a public cloud, and/or a private cloud. In this example, computerincludes a processor set(including processing circuitryand a cache), a communication fabric, a volatile memory, a persistent storage(including an operating systemand module, as identified above), a peripheral device set(including a user interface (UI), a device set, a storage, and an Internet of Things (IoT) sensor set), and a network module. Remote serverincludes a remote database. Public cloudincludes a gateway, a cloud orchestration module, a host physical machine set, a virtual machine set, and a container set.

Computermay take the form of a desktop computer, laptop computer, tablet computer, smart phone, smart watch or other wearable computer, mainframe computer, quantum computer or any other form of computer or mobile device now known or to be developed in the future that is capable of running a program, accessing a network and/or querying a database, such as remote database. As is well understood in the art of computer technology, and depending upon the technology, performance of a computer-implemented method may be distributed among multiple computers and/or between multiple locations. On the other hand, in this presentation of computing environment, detailed discussion is focused on a single computer, specifically computer, to keep the presentation as simple as possible. Computermay be located in a cloud, even though it is not shown in a cloud in. On the other hand, computeris not required to be in a cloud except to any extent as may be affirmatively indicated.

Processor setincludes one or more computer processors of any type now known or to be developed in the future. Processing circuitrymay be distributed over multiple packages, for example, multiple, coordinated integrated circuit chips. Processing circuitrymay implement multiple processor threads and/or multiple processor cores. Cacheis memory that is located in the processor chip package(s) and is typically used for data or code that should be available for rapid access by the threads or cores running on processor set. Cache memories are typically organized into multiple levels depending upon relative proximity to the processing circuitry. Alternatively, some, or all, of the cache for the processor set may be located “off chip.” In some computing environments, processor setmay be designed for working with qubits and performing quantum computing.

Computer readable program instructions are typically loaded onto computerto cause a series of operational steps to be performed by processor setof computerand thereby effect a computer-implemented method, such that the instructions thus executed will instantiate the method(s) specified in flowcharts and/or narrative descriptions of computer-implemented methods included in this document (collectively referred to as “the inventive methods”). These computer readable program instructions are stored in various types of computer readable storage media, such as cacheand the other storage media discussed below. The program instructions, and associated data, are accessed by processor setto control and direct performance of the inventive methods. In computing environment, at least some of the instructions for performing the inventive methods may be stored in modulein persistent storage.

Communication fabricis the signal conduction paths that allow the various components of computerto communicate with each other. Typically, this fabric is made of switches and electrically conductive paths, such as the switches and electrically conductive paths that make up busses, bridges, physical input/output ports and the like. Other types of signal communication paths may be used, such as fiber optic communication paths and/or wireless communication paths.

Volatile memoryis any type of volatile memory now known or to be developed in the future. Examples include dynamic type random access memory (RAM) or static type RAM. Typically, the volatile memory is characterized by random access, but this is not required unless affirmatively indicated. In computer, volatile memoryis located in a single package and is internal to computer, but, alternatively or additionally, the volatile memory may be distributed over multiple packages and/or located externally with respect to computer.

Persistent storageis any form of non-volatile storage for computers that is now known or to be developed in the future. The non-volatility of this storage means that the stored data is maintained regardless of whether power is being supplied to computerand/or directly to persistent storage. Persistent storagemay be a read-only memory (ROM), but typically at least a portion of the persistent storage allows writing of data, deletion of data and re-writing of data. Some familiar forms of persistent storage include magnetic disks and solid-state storage devices. Operating systemmay take several forms, such as various known proprietary operating systems or open source Portable Operating System Interface type operating systems that employ a kernel. The code included in moduletypically includes at least some of the computer code involved in performing the inventive methods.

Peripheral device setincludes the set of peripheral devices of computer. Data communication connections between the peripheral devices and the other components of computermay be implemented in various ways, such as Bluetooth connections, Near-Field Communication (NFC) connections, connections made by cables (such as universal serial bus (USB) type cables), insertion type connections (for example, secure digital (SD) card), connections made though local area communication networks and even connections made through wide area networks such as the Internet. In various embodiments, UI device setmay include components such as a display screen, speaker, microphone, wearable devices (such as goggles and smart watches), keyboard, mouse, printer, touchpad, game controllers, and haptic devices. Storageis external storage, such as an external hard drive, or insertable storage, such as an SD card. Storagemay be persistent and/or volatile. In some embodiments, storagemay take the form of a quantum computing storage device for storing data in the form of qubits. In embodiments where computeris required to have a large amount of storage (for example, where computerlocally stores and manages a large database) then this storage may be provided by peripheral storage devices designed for storing very large amounts of data, such as a storage area network (SAN) that is shared by multiple, geographically distributed computers. IoT sensor setis made up of sensors that can be used in Internet of Things applications. For example, one sensor may be a thermometer and another sensor may be a motion detector.

Network moduleis the collection of computer software, hardware, and firmware that allows computerto communicate with other computers through WAN. Network modulemay include hardware, such as a network interrace controller (NIC), a modem, software for packetizing and/or de-packetizing data for communication network transmission, and/or web browser software for communicating data over the internet. In some embodiments, network control functions and network forwarding functions of network moduleare performed on the same physical hardware device. In other embodiments (for example, embodiments that utilize software-defined networking (SDN)), the control functions and the forwarding functions of network moduleare performed on physically separate devices, such that the control functions manage several different network hardware devices. Computer readable program instructions for performing the inventive methods can typically be downloaded to computerfrom an external computer or external storage device through the hardware included in network module.

WANis any wide area network (for example, the Internet) capable of communicating computer data over non-local distances by any technology for communicating computer data, now known or to be developed in the future. In some embodiments, the WAN may be replaced and/or supplemented by local area networks (LANs) designed to communicate data between devices located in a local area, such as a Wi-Fi network. The WAN and/or LANs typically include computer hardware such as copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and edge servers.

End user device (EUD)is any computer system that is used and controlled by an end user (for example, a customer of an enterprise that operates computer), and may take any of the forms discussed above in connection with computer. EUDtypically receives helpful and useful data from the operations of computer. For example, in a hypothetical case where computeris designed to provide a recommendation to an end user, this recommendation would typically be communicated from network moduleof computerthrough WANto EUD. In this way, EUDcan display, or otherwise present, the recommendation to an end user. In some embodiments, EUDmay be a client device, such as thin client, heavy client, mainframe computer, desktop computer and so on.

Remote serveris any computer system that serves at least some data and/or functionality to computer. Remote servermay be controlled and used by the same entity that operates computer. Remote serverrepresents the machine(s) that collect and store helpful and useful data for use by other computers, such as computer. For example, in a hypothetical case where computeris designed and programmed to provide a recommendation based on historical data, then this historical data may be provided to computerfrom remote databaseof remote server.

Public cloudis any computer system available for use by multiple entities that provides on-demand availability of computer system resources and/or other computer capabilities, especially data storage (cloud storage) and computing power, without direct active management by the user. Cloud computing typically leverages sharing of resources to achieve coherence and economies of scale. The direct and active management of the computing resources of public cloudis performed by the computer hardware and/or software of cloud orchestration module. The computing resources provided by public cloudare typically implemented by virtual computing environments that run on various computers making up the computers of host physical machine set, which is the universe of physical computers in and/or available to public cloud. The virtual computing environments (VCEs) typically take the form of virtual machines from virtual machine setand/or containers from container set. It is understood that these VCEs may be stored as images and may be transferred among and between the various physical machine hosts, either as images or after instantiation of the VCE. Cloud orchestration modulemanages the transfer and storage of images, deploys new instantiations of VCEs and manages active instantiations of VCE deployments. Gatewayis the collection of computer software, hardware, and firmware that allows public cloudto communicate through WAN.

Some further explanation of virtualized computing environments (VCEs) will now be provided. VCEs can be stored as “images.” A new active instance of the VCE can be instantiated from the image. Two familiar types of VCEs are virtual machines and containers. A container is a VCE that uses operating-system-level virtualization. This refers to an operating system feature in which the kernel allows the existence of multiple isolated user-space instances, called containers. These isolated user-space instances typically behave as real computers from the point of view of programs running in them. A computer program running on an ordinary operating system can utilize all resources of that computer, such as connected devices, files and folders, network shares, CPU power, and quantifiable hardware capabilities. However, programs running inside a container can only use the contents of the container and devices assigned to the container, a feature which is known as containerization.

Private cloudis similar to public cloud, except that the computing resources are only available for use by a single enterprise. While private cloudis depicted as being in communication with WAN, in other embodiments a private cloud may be disconnected from the Internet entirely and only accessible through a local/private network. A hybrid cloud is a composition of multiple clouds of different types (for example, private, community or public cloud types), often respectively implemented by different vendors. Each of the multiple clouds remains a separate and discrete entity, but the larger hybrid cloud architecture is bound together by standardized or proprietary technology that enables orchestration, management, and/or data/application portability between the multiple constituent clouds. In this embodiment, public cloudand private cloudare both part of a larger hybrid cloud.

The instructions of moduleare now discussed with reference to the flowchart of, which illustrates a methodfor automatically compiling an improved answer to a question and providing it to the user, in accordance with an embodiment.

Steps of methodmay either be performed in the order they are presented or in a different order (or even in parallel), as long as the order allows for a necessary input to a certain step to be obtained from an output of an earlier step. In addition, the steps of methodare performed automatically (e.g., by computerof, or by any other applicable component of computing environment), unless specifically stated otherwise.

In a step, data may be received, including: a question provided by a user, a list that constitutes a direct answer to the question, and an introductory text to the list. The list and the introductory text may be generated or otherwise obtained by a conventional machine learning-based QA system. Module(of) may be part of such QA system, or the module may communicate with a separate QA system, for example using an API (Application Programming Interface) of that separate QA system.

The question may be provided in natural language, and may be phrased either explicitly as a question (for example, “What are the five tallest mountain peaks in the world?”) or only implicitly as a question (for example, “Types of user interfaces”). Both options may call for an answer in the form of a list-either a list of mountains in response to the first exemplary question, or a list of user interface types in response to the second exemplary question.

The list, as mentioned, constitutes a direct answer to the question—an answer which may entirely satisfy the user's need for information. This stands in contrast to traditional search results, that typically only include a set of hyperlinks to Web pages that are deemed relevant to a user's query, along with an excerpt from each Web page; the user then has to follow one or more of these hyperlinks in order to read the information provided on the pertinent Web pages and potentially find an answer (or a number of alternative answers) to the question.

For example, the list which may be obtained by the QA system in response to the question “What are the five tallest mountain peaks in the world?” may be:

The introductory text may include one or a few sentences generally describing the contents of the list. In many conventional QA systems, the introductory text is presented immediately before the list. The introductory text may or may not convey information beyond what is already included (explicitly or implicitly) in the question itself. For example, some introductory texts retrieved or generated by QA systems merely paraphrase the question; for instance, for the question “What are the five tallest mountain peaks in the world?,” the introductory text may be “Top 5 highest mountains in the world.” Such introductory text may be considered redundant, because the list provides the user with exactly the answer he or she were seeking. In contrast, some introductory texts may convey information that is not included in the question itself, and is necessary for the user to correctly understand the list; for instance, in response to the question “Eligibility for a green card,” the introductory text may be “Eligibility Criteria for a Green Card through Family,” meaning that the appended list only includes the criteria for obtaining a Green Card through family ties, and not, for example, through employment or asylum. Such introductory text is certainly nonredundant, because its associated list does not precisely or fully addresses the user's question; without that introductory text, the user may falsely believe that the list is an exhaustive list of all the criteria for obtaining a Green Card, when in fact the list only constitutes a very partial answer to the question.

Another scenario in which an introductory text may be deemed redundant is when the list itself coveys the same (or approximately the same) information as the introductory text. For example, assume that the question is “How to replace a car tire,” the introductory text is “Steps to replace a car tire,” and the items of the list are of the style: “Step 1: Loosen the bolts . . . ”, “Step 2: Use a jack stand to . . . ”, etc. Clearly, the words “Step X” in every list item make the introductory text redundant.

In a step, a machine learning model (or “model” for short) may be used to classify the introductory text as redundant or nonredundant, based on the data received in step. Namely, the question, list, and introductory text may be provided as inputs to the model, such that the model outputs a class name (be it “redundant”/“nonredundant” or any other names or even unique numbers conveying the same intent) which most probably suits the introductory text.

The model is optionally a language model, but could be any suitable machine learning model capable of classifying texts of the kind of a typical introductory text into two classes.

The model may undergo training prior to execution of method, for example using the following steps:

First, training data may be obtained, such as from prior executions of a QA system. The training data may include multiple samples, each including: a question provided by a user, a list generated by the QA system in response to the question, and an introductory text generated by the QA system in response to the question. In experiments performed by the inventors, such training data was obtained from the following three publicly-available datasets: NQ (Tom Kwiatkowski et al., “Natural Questions: A Benchmark for Question Answering Research,” in Transactions of the Association of Computational Linguistics, 2019), GooAQ (Daniel Khashabi et al., “GooAQ: Open Question Answering with Diverse Answer Types,” in Findings of the Association for Computational Linguistics: EMNLP 2021, pp 421-433), and CCQA (Patrick Huber et al., “CCQA: A New Web-Scale Question Answering Dataset for Model Pre-Training,” in Findings of the Association for Computational Linguistics: NAACL 2022, pp 2402-2420).

Then, for each of these samples, a similarity measure may be computed between the respective question (optionally, together with the list) and the respective introductory text, to quantify the degree of similarity between the two. For example, the similarity measure may be in the form of an overlap score S, computed according to the following formula: